Source code for spack.directives_meta

# Copyright Spack Project Developers. See COPYRIGHT file for details.
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)

import collections.abc
import functools
from typing import Any, Callable, Dict, List, Optional, Sequence, Set, Type, Union

import llnl.util.lang

import spack.error
import spack.repo
import spack.spec

#: Names of possible directives. This list is mostly populated using the @directive decorator.
#: Some directives leverage others and in that case are not automatically added.
directive_names = ["build_system"]


[docs] class DirectiveMeta(type): """Flushes the directives that were temporarily stored in the staging area into the package. """ # Set of all known directives _directive_dict_names: Set[str] = set() _directives_to_be_executed: List[Callable] = [] _when_constraints_from_context: List[spack.spec.Spec] = [] _default_args: List[dict] = [] def __new__( cls: Type["DirectiveMeta"], name: str, bases: tuple, attr_dict: dict ) -> "DirectiveMeta": # Initialize the attribute containing the list of directives # to be executed. Here we go reversed because we want to execute # commands: # 1. in the order they were defined # 2. following the MRO attr_dict["_directives_to_be_executed"] = [] for base in reversed(bases): try: directive_from_base = base._directives_to_be_executed attr_dict["_directives_to_be_executed"].extend(directive_from_base) except AttributeError: # The base class didn't have the required attribute. # Continue searching pass # De-duplicates directives from base classes attr_dict["_directives_to_be_executed"] = [ x for x in llnl.util.lang.dedupe(attr_dict["_directives_to_be_executed"]) ] # Move things to be executed from module scope (where they # are collected first) to class scope if DirectiveMeta._directives_to_be_executed: attr_dict["_directives_to_be_executed"].extend( DirectiveMeta._directives_to_be_executed ) DirectiveMeta._directives_to_be_executed = [] return super(DirectiveMeta, cls).__new__(cls, name, bases, attr_dict) def __init__(cls: "DirectiveMeta", name: str, bases: tuple, attr_dict: dict): # The instance is being initialized: if it is a package we must ensure # that the directives are called to set it up. if cls.__module__.startswith(spack.repo.ROOT_PYTHON_NAMESPACE): # Ensure the presence of the dictionaries associated with the directives. # All dictionaries are defaultdicts that create lists for missing keys. for d in DirectiveMeta._directive_dict_names: setattr(cls, d, {}) # Lazily execute directives for directive in cls._directives_to_be_executed: directive(cls) # Ignore any directives executed *within* top-level # directives by clearing out the queue they're appended to DirectiveMeta._directives_to_be_executed = [] super(DirectiveMeta, cls).__init__(name, bases, attr_dict)
[docs] @staticmethod def push_to_context(when_spec: spack.spec.Spec) -> None: """Add a spec to the context constraints.""" DirectiveMeta._when_constraints_from_context.append(when_spec)
[docs] @staticmethod def pop_from_context() -> spack.spec.Spec: """Pop the last constraint from the context""" return DirectiveMeta._when_constraints_from_context.pop()
[docs] @staticmethod def push_default_args(default_args: Dict[str, Any]) -> None: """Push default arguments""" DirectiveMeta._default_args.append(default_args)
[docs] @staticmethod def pop_default_args() -> dict: """Pop default arguments""" return DirectiveMeta._default_args.pop()
[docs] @staticmethod def directive(dicts: Optional[Union[Sequence[str], str]] = None) -> Callable: """Decorator for Spack directives. Spack directives allow you to modify a package while it is being defined, e.g. to add version or dependency information. Directives are one of the key pieces of Spack's package "language", which is embedded in python. Here's an example directive: .. code-block:: python @directive(dicts='versions') version(pkg, ...): ... This directive allows you write: .. code-block:: python class Foo(Package): version(...) The ``@directive`` decorator handles a couple things for you: 1. Adds the class scope (pkg) as an initial parameter when called, like a class method would. This allows you to modify a package from within a directive, while the package is still being defined. 2. It automatically adds a dictionary called "versions" to the package so that you can refer to pkg.versions. The ``(dicts='versions')`` part ensures that ALL packages in Spack will have a ``versions`` attribute after they're constructed, and that if no directive actually modified it, it will just be an empty dict. This is just a modular way to add storage attributes to the Package class, and it's how Spack gets information from the packages to the core. """ global directive_names if isinstance(dicts, str): dicts = (dicts,) if not isinstance(dicts, collections.abc.Sequence): message = "dicts arg must be list, tuple, or string. Found {0}" raise TypeError(message.format(type(dicts))) # Add the dictionary names if not already there DirectiveMeta._directive_dict_names |= set(dicts) # This decorator just returns the directive functions def _decorator(decorated_function: Callable) -> Callable: directive_names.append(decorated_function.__name__) @functools.wraps(decorated_function) def _wrapper(*args, **_kwargs): # First merge default args with kwargs kwargs = dict() for default_args in DirectiveMeta._default_args: kwargs.update(default_args) kwargs.update(_kwargs) # Inject when arguments from the context if DirectiveMeta._when_constraints_from_context: # Check that directives not yet supporting the when= argument # are not used inside the context manager if decorated_function.__name__ == "version": msg = ( 'directive "{0}" cannot be used within a "when"' ' context since it does not support a "when=" ' "argument" ) msg = msg.format(decorated_function.__name__) raise DirectiveError(msg) when_constraints = [ spack.spec.Spec(x) for x in DirectiveMeta._when_constraints_from_context ] if kwargs.get("when"): when_constraints.append(spack.spec.Spec(kwargs["when"])) when_spec = spack.spec.merge_abstract_anonymous_specs(*when_constraints) kwargs["when"] = when_spec # If any of the arguments are executors returned by a # directive passed as an argument, don't execute them # lazily. Instead, let the called directive handle them. # This allows nested directive calls in packages. The # caller can return the directive if it should be queued. def remove_directives(arg): directives = DirectiveMeta._directives_to_be_executed if isinstance(arg, (list, tuple)): # Descend into args that are lists or tuples for a in arg: remove_directives(a) else: # Remove directives args from the exec queue remove = next((d for d in directives if d is arg), None) if remove is not None: directives.remove(remove) # Nasty, but it's the best way I can think of to avoid # side effects if directive results are passed as args remove_directives(args) remove_directives(list(kwargs.values())) # A directive returns either something that is callable on a # package or a sequence of them result = decorated_function(*args, **kwargs) # ...so if it is not a sequence make it so values = result if not isinstance(values, collections.abc.Sequence): values = (values,) DirectiveMeta._directives_to_be_executed.extend(values) # wrapped function returns same result as original so # that we can nest directives return result return _wrapper return _decorator
[docs] class DirectiveError(spack.error.SpackError): """This is raised when something is wrong with a package directive."""